The risks of industry influence in tech research

Summary

This Perspective argues that the technology industry exerts pervasive and underappreciated influence over the scientific study of its own products, paralleling historical interference by tobacco, fossil fuel, and soda industries but amplified by unique asymmetries: tech firms control the data, funding, and platform access required to study social media, search, and AI. The authors contend that existing scientific safeguards — IRB review, conflict-of-interest disclosures, and Open Science norms — are inadequate and can even disadvantage independent researchers relative to industry-affiliated ones. They catalog six mechanisms of influence and propose reforms spanning journals, funders, universities, and governments, including mandated data access and an IPCC-style intergovernmental panel for information technology.

Key Contributions

• A unified taxonomy of industry influence mechanisms: burying internal research, selective publishing, design bias, selective funding, inhibition of independent research, and performative collaboration.
• Extension of agnotology and “merchants of doubt” frameworks (Oreskes, Proctor, O’Connor, Holman & Bruner) into platform studies and computational social science.
• Diagnosis of how Open Science reforms (preregistration, large samples, RCTs) can be co-opted or asymmetrically applied to favor industry-produced work.
• Concrete policy proposals: mandatory IRB review for industry research, stronger COI auditing, mandated data access building on the EU DSA, disclosure of internal industry studies, dedicated independent funding streams, and an Intergovernmental Panel on Information Technology.
• Practical heuristics for academics evaluating industry collaborations, including identifying design bias, “first-peek” risks, and data-integrity concerns.

Methods

A conceptual Perspective synthesizing prior literature on industry-science conflicts and applying it to tech. The argument is built through case-based analysis of historical industries (tobacco, asbestos, fossil fuels, soda) and contemporary tech cases (Meta, X/Twitter, Google, Coca-Cola), examination of leaked internal documents (Haugen disclosures, 2025 Meta whistleblower cache), comparison of statistical choices across Meta-affiliated publications, and an informal OpenAlex bibliometric check of Meta-funded researchers versus disclosed Meta funding.

Findings

• Internal research at Meta and other firms has been suppressed or deleted on topics including harms to children, teen mental health, and VR safety.
• Industry-affiliated studies apply lenient statistical thresholds (e.g., p<.1) when measuring benefits but stricter multiple-comparison corrections when measuring harms, producing asymmetric conclusions.
• RCTs of short platform “breaks” systematically underestimate long-term, cumulative, and societal-scale harms.
• Independent data access has eroded sharply: CrowdTangle shutdown, X API restrictions, legal threats against NYU Ad Observatory and CCDH.
• The Meta/Social Science One collaboration showed design bias, undisclosed platform interventions during study windows, missing user data, and deviations from preregistration.
• Massive under-disclosure of industry ties: ~1000+ Meta-funded researchers but only 472 OpenAlex publications acknowledging Meta funding.
• Industry research is frequently exempted from IRB review, blocking independent replication.

Connections

This paper is a foundational critical lens for the platform-governance-and-data-access cluster, and pairs directly with Bak-Coleman2026-mk as companion work by the lead author. Its concerns about eroded API access, performative collaboration, and asymmetric data availability resonate strongly with empirical and methodological critiques in Freelon2024-sc, Rieder2025-ju, Rieder2026-pp, Murtfeldt2025-wu, and Donovan2025-ws, as well as with DSA-implementation analyses such as Vincent_undated-re and Schiffrin_undated-gi. The critique of industry-shaped causal inference (e.g., short-break RCTs) speaks to debates over evidentiary standards in works like Allen2025-ot and Lewandowsky2026-ob.